Discrete blind reconstruction method for multi-coset sampling

Abstract

Multi-coset sampling is a compressed sampling strategy for the acquisition of spectrally sparse signal. In previous discrete blind recovery methods, either the sensing matrix is really huge, or the sampled sequences are interpolated to the Nyquist rate at the first step, both leading to a high computational complexity. A new discrete blind reconstruction method is proposed in this study to reconstruct a multiband signal from its multi-coset samples. In the proposed method, the fractional delay is implemented at sub-Nyquist rate, so that the sampled sequences do not need to be interpolated to the Nyquist rate until the final step of the reconstruction process. Hence, the computational complexity of the reconstruction method is reduced. Moreover, a windowing procedure is employed in this method to avoid the spectral spreading effect. Numerical experiments are presented to demonstrate that the authors’ method outperforms previous methods in terms of computational complexity and accuracy.